Hacker News Reader

Comment by snakeyjake | Hacker News Reader

> Making USB reversible to begin with would have necessitated twice as many wires and twice as many circuits, and would have doubled the cost.

This is... not correct.

You can make USB reversible with 1 extra pin and 1 extra wire. Grounds on pins 1 and 5, data on pins 2 and 4, and VCC on pin 3. Then have those pins on both sides of the plug and a socket with a single set of contacts on one side.

That's BASICALLY what Apple did with lightning.

Then you implement auto crossover detection, (edit: Gah! you don't even have to do that just flip the flipping wires) which had been around for years and is dirt cheap, in the hub. It would have been like, six, more transistors in the hub IC.

edit: I completely forgot that reversible USB 2.0 plugs already exist and use a simpler (and cheaper) method. They just tend not to be so reliable because of the thinner materials and the fact that they're not spec-compliant so they tend to be grey market jobs made for the lowest price possible.

Here is one: https://www.amazon.com/Tripp-Lite-Universal-Reversible-UR050...

No doubling of wires or circuits required, just a thin double-sided PCB.

Was the connector form-factor inherited from an earlier project and the players didn't want to design a new one?

View on HN

There are also reversible USB plugs that are a single "tongue", inheriting that design from USB drives that don't have any plug to speak of but look more like a card-edge connector:

https://www.pcgamer.com/youre-telling-me-we-could-have-had-r...

https://2b.com.eg/en/2b-cv177-cable-usb-type-a-reversible-pl...

I got these by accident, spent a month thinking I had incredible USB skills, then discovered and became fascinated by the tongue.

This is like in basketball where they throw it right next to the hoop for someone to make an innuendo slam dunk, right??

PCgamer:

> Incidentally, the late Steve Jobs deserves a tip-of-the-hat for helping USB become a long-lasting standard. It was in 1998 that USB made some real headway, courtesy of the iMac G3, the first computer to ship with only USB ports for external devices (there were no serial or parallel ports). That came three years after USB 1.0 debuted, with a data rate of 12Mbps.

Also PCgamer:

> Given that it was a protocol designed to counter the typically Apple expense of sticking Firewire onto a motherboard, cost was the biggest reason for the eventual success of USB as a thing

Apple started building FireWire 400 into Macs when 12Mbps USB 1.0 was not really an equivalent. USB 2.0 didn't make the cut for the first generation iPod but was supported on subsequent models.

> the first computer to ship with only USB ports

I remember presenting this in a project as an undergraduate and I was challenged by the lecturer that it would be highly unusual for a consumer device to ship without RS232 and I could only say “yeah”

> I remember presenting this in a project as an undergraduate and I was challenged by the lecturer that it would be highly unusual for a consumer device to ship without RS232 and I could only say “yeah”

It's kinda funny that this makes complete sense to me and I wouldn't have argued it either, but at the same time despite having been using computers since the late '80s the only thing I've ever had that required the use of a serial port for normal operation was my 2001-era Palm m100.

PS/2 keyboards/mice, parallel printers, and gameport controllers I had many of each, but all my modems were internal so the first time I ever plugged something in to a serial port was a null-modem cable in late '99 so my friend and I could use Windows 98 Direct Cable Connection to play Counter-Strike against each other without having to buy ethernet cards.

Aside from that one Palm device everything I've ever plugged in to an actual serial port has been a temporary thing like that null-modem connection. Mostly serial console for last resort access to misbehaving network appliances. Even then, most devices that have had the slightest bit of effort put in to them in the last 20 years have at least built in a USB-Serial adapter these days so we don't have to deal with the battle of devices that want full 12v RS232 and USB-DE9 adapters that deliver somewhere between 5 and 8.

To be fair for very long time most motherboards had RS232 port on headers, just not on back plate.

Some pro workstation boards still do!

WX-X299-PRO or its newer cousin ASUS Pro WS W680-ACE

The problem with USB-a isn't that it's non reversible, it's that it's non-reversible _and_ rectangular, so it's not clear at a glance which way round it should go.

All they had to do was make the connector have a non-symmetrical shape so that it's immediately obvious which way round it goes when you pick it up - you could do it without even looking. Think of how much time we'd have collectively saved with this minor design change.

That only works when you can see the place you're connecting to, or have used it many times.

I've definitely cursed many times failing to plug in non-rectangular VGA, serial and parallel cables ;)

[edit: it is still better, but not a panacea]

Even having a bump on the rubber moulding as part of the standard would have solved the problem.

The USB symbol goes on top, IIRC.

Edit: ninja'd by an enormous amount of people elsewhere in the thread, gah.

This is not always true, and not all ports are horizontal anyway.

Isn't it part of the standard?

In relation to the PCB, so if they're horizontally mounted onto the motherboard, you end up with vertical slots on the back of a standard tower.

Yes.

Does everyone follow the standard? No. Including big names that prefer to have their logo facing up.

Apple and many others need to be reverse always. There is no top.

Vertically aligned motherboards are hard to judge what "up" is. Usually up is where the expansion boards are plugged in. But not all computer users know how a computer looks inside.

> The USB symbol goes on top, IIRC.

For some values of "top"; that is, some device manufacturers install the socket upside-down.

It doesn't help if you've trying to do it totally blind, but the 2 open square holes go on top, once I figured that out I've never had any issues plugging them in correctly. The only one I have issues with is my keyboard because the pcb in the plug is black and the it's a little hard to tell which side is the open holes and which side is the black pcb.

Heck, just making the plastic insert white instead of black (now sometimes blue, occasionally orange) would have helped.

Frequently (by convention) the colors on the USB ports have meaning (including white actually indicating USB 1.0):

https://duckduckgo.com/?q=colors+on+usb+ports

> All they had to do was make the connector have a non-symmetrical shape so that it's immediately obvious which way round it goes when you pick it up - you could do it without even looking. Think of how much time we'd have collectively saved with this minor design change.

I disagree. When connecting an HDMI cable I sometimes have issues, especially if in a weird angle.

However I do concede is faster to connect an HDMI than a USB

Yeah, I want the connector that can be attached by only moving the TV a few inches, just enough to get my hand to fit and feel around, not something I have to rearrange the furniture for and break out some headlight to see which direction the cable is oriented. With BNC cables, I could do it with my eyes closed.

I find that a USB-A fits reasonably by feel into an Ethernet port for extra points.

Oh good, and here I thought I was the only one who had done that... more than once.

Admittedly I did this today.

I did that today! right hand side of lenovo X230, then wondered why my mouse didn't work

Conversely, I plugged my USB-C cable into a USB-A port by feel recently, and didn’t notice until much later that it wasn’t charging.

Tbh, humanity should've just used SDI for everything.

The problem with the HDMI connector (well, the problem on topic here) is that it's almost rectangular. Just a pair of small chamfers doesn't cut it.

Display port has a much more usable design, with a chamfer that is about twice the size as HDMI.

For an example of an asymmetrical "rectangle", look at HDMI: https://en.wikipedia.org/wiki/HDMI#/media/File:HDMI_connecto...

DisplayPort is even simpler; a notched shaved off of one of the corners.

Still very difficult to plug in, also often struggling with its orientation.

FireWire was great for that. I loved that connector.

Wasn't it based on the GB link cable?

No, they both used the same physical form factor, but it seems to have just been a model from a OEM used by both nintendo and apple around that time, rather than one spec deliberately copying the connector from the other.

I’ve seen that claim but AFAICT it comes from a single source, which was a presentation by someone involved in the FireWire connector design. So at this point it’s more of an urban legend, until we have some solid confirmation.

It does not really matter though, they are very similar in construction, and are very sturdy, and I really like them both :)

I've seen the same claim made. Apparently it was the sturdiness that led to it being copied. Good enough for kids beating on it and stepping on the ends, more than good enough for professional applications.

The 6-pin connector was nice. But laptops and camcorders generally had the 4-pin connector (sans power)...not so nice.

Indeed. The small one was very flimsy and annoying to use.

They did that from the beginning with USB-B, but they never standardized an orientation.

Just curious if you have seen the special irregular USB A plug for the Lenovo Yoga 3 Pro. If so, what are your thoughts on it?

https://ae01.alicdn.com/kf/HTB1XeKTDHSYBuNjSspiq6xNzpXae.jpg...

imgur mirror https://i.imgur.com/RQvMfhT.png

random newegg link: https://www.newegg.com/p/2MV-00MD-00DX3

Right, don't make a non-reversible connector appear reversible... I would have thought this is Connector 101.

It's so bad it takes me 3 attempts to plug in USB-A blind, at least 50% of the time.

other part of connector 101 is "put the springy bit on the cable, not socket, so when it wears out you replace a cable, not connector"

Given how Mini-USB looked I'm gotta say nobody there had any clue about designing connectors...

Right, the only thing worse than USB was the PS2 port. What sort of madness drive someone to invent a round connector that is keyed.

Hyperbole aside, there are some very good reasons to go with round connectors, it can be very strong and you can put a big nut on them to have an extremely secure lock. however the PS2 connector is nether of these things, It would require a much more robust key to work well.

The worst connector I ever encountered was on a old ATI capture card, it was the same barrel as a ps2 connector but had about 15 pins. Everything had to be perfect for it to insert, a tiny scratch or a slightly bent pin, it was not going anywhere.

The PS/2 connector was just a mini version of DIN connectors from the mid 1900s. They were sometimes used for printers and floppy drives in place of wider parallel connectors. In general, once you knew where ‘up’ was you could figure out the orientation in a way that is still more intuitive than USB. XLR cables are similar and still widely used, but have more external cues as to their orientation.

https://en.m.wikipedia.org/wiki/DIN_connector

> Right, the only thing worse than USB was the PS2 port. What sort of madness drive someone to invent a round connector that is keyed.

The Apple desktop bus (ADB) that was used for keyboards and mice in Mac-land was also round. They improved it by using asymmetrical plugs, which marginally helped.

I would line up the barrel and then rotate while pushing in until it plugged in. I don’t remember bending any pins. The mark on the top helped until the tower form factor turned the motherboard sideways. Still do this for XLR and MIDI but they’re more robust.

I guess it would be easier with the larger size of the USB-A connector, but I still struggle to plug in micro-USB cables sometimes, as it's still difficult to figure out its orientation at a glance.

Tbf I think the most prevalent issue with USB was from people plugging them in where they cannot see the port properly, ie the back of a computer tower (most often) or back of TV/console etc.

Not really sure that that's USB's fault so much as a lack of front USB ports. I think in those cases the consumer can chose to get some sort of hub that's easier to access/can be pulled out from behind their setup.

Now that C exists though, idk if we really need to go much further. Connector much smaller than it is now will become fragile.

I've always thought they could add a single fiber core to USB cables for some "ultra high speed" standard, since USB4 is predicted to top out at 80GBps, but we could get 50Tbps or so over a single fiber if we really needed to.

But after owning several fiber cables, for networking and HDMI fiber based cables...yeah they're far, far too weak to replace copper. Just look at 'em wrong and they snap/break.

Plus tbf if we get 80gbps from usb4...the problem then becomes that we don't reaaaally have a use case for that speed. Maybe network adaptors, but beyond that...I mean a modern SSD will saturate that with a sequential read I suppose, but still.

I think he DID explain why: he didn’t and doesn’t know how

To help visualize:

    1: GND
    2: DATA +/-
    3: VCC
    4: DATA +/-
    5: GND

With that pinout you'd end up with one orientation being full/high-speed and the other being low speed.

Edit: see https://en.wikipedia.org/wiki/USB_communications#Signaling_s...

Would be an acceptable tradeoff.

But when the plug was designed, there was only one speed.

Only if you don't also redefine the convention to select speeds.

Can't that square metal be used as ground?

You don’t always want the shield coupled to the signal ground. If it’s coupled on both sides then you can have all kinds of fun problems. Ideally, the shield should be connected to the ground on one end

> Ideally, the shield should be connected to the ground on one end

Depends on the signal. With low-frequency stuff (like audio) you want to ground the shield on one side to avoid ground loops, but with high-frequency stuff (USB 3, for example) you definitely want to ground the shield on both sides: if you don't, you end up creating an antenna which is really bad for your EMI performance.

USB 1.x happens to be at an in-between point where neither issue is a big deal. The connector spec literally says "use accepted industry practices", so either approach is fine if it manages to pass emissions testing.

That goes against all the guidance I’ve read, but I’m an amateur so you may be right. I’m confused on why it relates to signal speed though. For instance, I’ve read that this guidance also applies to Ethernet which is quite high frequency and based on differential pairs

> Question: “I have been told that you should only ground the Ethernet cable at one end. Is this true?”

> Answer: “Sometimes yes, but only if you are seeking to avoid a ground loop. You likely don’t have to worry about this. If you are running shielded Ethernet between two structures that have their own separate AC systems or separate grounding rods installed, then YES, you should worry about what is known as a ground loop. A ground loop occurs when you have actual conflicting AC ground systems. Merely bonding your Ethernet cable at multiple points to the same AC grounding system in a single structure does not create ground loops. Think in terms of ground rods. How many ground rods are involved? If you have more than one with more than one structure involved then you are wise to bond to ground at a single end only.”

From https://www.truecable.com/blogs/cable-academy/residential-bo...

> Then you implement auto crossover detection, dwhich had been around for years

Mdi-x/auto crossover for ethernet was introduced in 1999, that is after USB (1996) was designed, not sure of the technique was known at that time

There were auto-crossover RS-232 boxes going back deep into the 1980s.

with baud rate detection to boot (down the line)

You can't do "auto crossover" with USB because it needs to distinguish between D+/D- for speed selection.

https://electronics.stackexchange.com/questions/73295/in-a-u...

It works for Ethernet because the transmit and receive lanes are separate, but with USB, there's only one pair used for both directions.

It doesn't need to.

It chose to. They could just use different way to do it. Like, I dunno, a single packet of negotiation in slow mode...

Not really relevant.

Ethernet needs crossover because it has dedicated transmit/receive pairs, but USB 1.x/2.0 uses a single wire pair for both tx & rx. Ethernet crossover is pretty much identical to serial null modem cables, and those were around for a few decades by then.

USB "crossover" of the single differential wire pair would be an inversion of 0 and 1. Pretty trivial to do once you figure out a way to communicate crossover or not.

Ugh, when I read that line I immediately came here to comment, but you beat me to it.

I really really really hope that this was just an error by the journalist, and not Bhatt essentially lying to justify the choice.

Couldn't original USB have used TRRS? That's better than reversible - an infinite number of orientations are allowed.

That shorts together a lot of the contacts as you're plugging it in.

There are of course various devices which somehow ended up using TRRS for USB, but unplugging/plugging those while powered is definitely not recommended:

https://www.reddit.com/r/ErgoMechKeyboards/comments/rt083u/a...

http://www.totekinternational.com/35mm-trrs-to-usb-a-male-ca...

https://pinoutguide.com/PortableDevices/ipod_shuffle_pinout....

Most of this is just cheap design on part of the ergomech community. It's not that hard to add some protection. E.g. some keyboards use a Schottky diode on VCC for reverse polarity protection and ESD protection diodes to protect the data lines:

https://github.com/zsa/docs/blob/master/Ergodox%20V5-L.pdf

Of course the design requirements of USB would be different, since you'd always plug/yank the cable on a machine that is on, rather than protect against doing this accidentally sometime. Given that, it probably made more sense to completely avoid the issue.

It's more an issue of ESD protection being kinda tricky on DIY boards due to the need to stick with THT parts. When you're building a $200-$300 keeb, a $0.05 diode doesn't really all that much price-wise.

ESD protection diodes also don't really solve the problem. They are intended for short-duration high-voltage low-current spikes, not long-duration low-voltage high-current ones. Definitely better than nothing, though!

I'm not an expert but couldn't you just make the tip live and not connect the rest until that's connected? Or have a switch for when the plug is fully inserted. I see that it's a problem for sure, but it seems easier than the problems they solved to make usb c reversible

> I'm not an expert but couldn't you just make the tip live and not connect the rest until that's connected?

No. You have to deal with disconnects being possible at both the host side and the device side. Making the tip live means you're touching a live wire to each contact in turn while plugging in the device side.

They'd have to make it not fit in a normal TRRS hole or use some fancy power negotiation stuff to avoid blowing up headphones. It would be cool though. But perhaps easier to mess up than USB. Cheap headphone cables aren't great.

TR/TRRS have very small contact patches since you're connecting to the tangent of a circle. That's not good for anything carrying more than a very small current. And even then, it can be an unreliable contact.

If you want to do it properly:

    1: GND/sense
    2: DATA +/-
    3: VCC
    4: DATA +/-
    5: GND/sense

Have the host side connect only pin 1 to GND and leave pin 5 floating, but have the device side connect both 1&5 to gnd, with individual pullups for 1 & 5 to VCC.

The device will sense LO on 1 and HI on 5 if either both sides are plugged in straight or both sides are plugged in reversed, but it will sense HI on 1 and LO on 5 if exactly one of them is reversed. This allows the device to detect when inversion of D+/D- is needed.

>edit: Gah! you don't even have to do that just flip the flipping wires)

Isn't that essentially what all ethernet cables do? it's not like this something never done before.

When was Auto MDI-X introduced? A quick google failed to get me an answer but I remember USB being available while I still needed a crossover cable to hook up 2 switches.

oh god, I was made to be an idiot when I once told someone they would need a crossover cable to do what they wanted to do a few years ago...

If it makes you feel any better... for some reason there are some currently in production business jets that still require crossover cables for maintenance... so they ARE still required sometimes.

I've also dug further into my parts bin thinking "oh, that yellow cable would've been perfect, but I need a non-crossover cable"... could've used the yellow one instead of digging further. Ah well.

Security through obscurity?

You encounter this in older industrial equipment. Some of the old ethernet chipsets don't support automatically switching the pairs around. It's not really a problem since at least in our local area they need to be installed by licensed cablers who are used to dealing with legacy devices.

Many devices won't get updated until chipsets go EOL or there's some other compelling reason to do so.

Having said that, modern chipsets seem to all support it.

Together with 1Gbit ratification, so ~2000

Ethernet hardware now has auto-crossover. However, that hasn't always been the case. There was a good decade in the 90's where you needed a special cable wiring if you wanted to connect two computers vs a computer to a switch.

Hah! I recently had to put ends on a Cat 6 cable for weird reasons, and spent a few minutes deciding if it needed to be crossover or not.

The worst is trying to repair one end of an existing cable and having to think really hard about whether to use A or B wiring.

sometimes, it's easier to replace both ends. or replace the same end, twice

Random anecdote: I worked at a company with an on-prem server room and the building got struck by lightning. It fried some servers and quite a few nic cards on employee computers. Our CTO decided since we'd be in the building doing hardware replacements anyway, now was the time to redo our wiring closest as well.

That was my introduction to wiring up CAT-5 cables and rj45 connectors and, with so many to make, I needed a way to remember the wiring diagram without having to read it over and over or use a visual reference. The mnemonic rhythm I used to remember the pattern was "orange/white, orange... green/white, blue... blue/white, green... brown/white, brown". Of course, said out loud, you couldn't tell if it was "orange/white, orange" or "orange, white/orange" and I kept confusing the rest of the team when I'd say it, but it worked for me.

This was circa 2005/2006 and I still say it like that. It just stuck.

Stripey orange

Orange

Stripey green

Blue

Stripey blue

Green

Stripey brown

Brown

Thanks for the memories!

I once used a crossover cable with an old 3Com card and a Linksys switch - whatever it sent down the wires, the Ethernet port on that switch was fried and has never done Tx or Rx since. I bandaged it with electric tape so I don't connect something else, and also in hope it will someday heal.

The 3Com card was unscathed. In fact, it is probably feeling stronger after that exertion.

Easily going into the early 2000s. I remember trying to setup LAN parties in 2004/2005, where the wrong combination of patch or normal cable with hub/switch/PC made everything a nightmare.

I found a few patch cables mixed in with regular cables. one day, i got so tired of messing with them, i cut the cables and put them in the trash. i could have just trashed, them, but that didn't feel right. i mean, what if someone were to go dumpster diving. the curse would then be on them.

Whenever I have a bum cable and confirm it’s the cable, it gets destroyed. Too many cables have failed to get into the trash or have somehow crawled back out for me to do anything else.

I've certainly more than once outright thrown out items for just falling outside of the return window rather than hand-me-downing them to friends for being so crappy that I didn't want to ever operate them ever again in any context.

The one I remember the most was a Bluetooth speaker that yapped your ear off for a half-minute combined with an obscene amount of jingle beeping at 100% blast and more LEDs than a Christmas tree as a power-on/off procedure back when they were novel items.

Uh? They're trivial to tell apart, just put the RJ45 jacks side by side and look at the colors of the wires going in?

Same sequence -> Straight ethernet cable

Same sequence except for 1 swapped pair -> Probably a crossover Ethernet cable

Inverted sequence -> ISDN cable, burn it.

Also (apart from the ISDN cables) - it hasn't mattered for close to quarter of a century, all switches and NICs are auto-MDI-X nowadays, why throw away the stuff.

Now I know I'm old because I just assumed that's how it still was. I'd be like the weird old grandpa at a party talking about a choke or carburetor or a hand crank on a car while all the kids just stare blankly.

You have to consider that initially it was supposed to be connector for stuff like mouses or keyboards or gamepads or printers. 12Mbits.

But yeah that was only one of many misdesigns along the way. Like making turning ports on and off optional features, mini-usb (seriously, none other usb version I've seen failed as often as mini-usb connectors), just how overcomplicated USB-PD is, or the recent mess in USB-C cabling.

Or just put + and - through a rectifier and you don't need an extra pin.

Data lines -- use differential signalling and polarity determined by whether +/- got reversed by the rectifier or not.

Or do like Comcast does and put all bidirectional data through 1 conductor at multi-gigabit speeds.

How are you going to implement the rectifier, though?

The usual implementation is using diodes, but that means you lose twice the diode forward voltage to the rectifier. Even with a Schottky diode that means you are losing 0.5V-1.0V. That's a 10%-20% loss at 5V!

You could use a transistor-based "ideal diode" solution, but that's not going to be cheap - especially in the 90s.

Maybe if you have no clue about electronics don't talk about electronics...

> Or do like Comcast does and put all bidirectional data through 1 conductor at multi-gigabit speeds.

And how much modem's transceiver cost ? And how does that compare to cost of computer mouse or other cheap USB peripherals.

They'll cost 1/100 the price if every USB port in the world uses them. Auto parts for common models of cars are so incredibly cheap for a reason.

USB1 was designed in times where transistors were significantly more expensive. And in times where laptops or phones or any other situation when you disconnected peripherals were far less common. Like, it still was improvement over PS/2.

It's also not just few transistors to flip the lines around, it's also the logic to drive that

Y'all are talking as if you were in the vacuum tube era.

There were 42 million transistors in a Pentium 4, I'm sure they could have put a few thousand per USB port if they thunk optimistically instead of pessimistically.

When USB came out we already had some very crude 3D realtime rendering already at that time available to mass consumers, I'm sure we could have figured out how to flip power on a port if we put our minds to it.

Think "how can we make shit happen" instead of excuses of why it's hard.

It's by doing hard things that we got to the 80 billion transistors of modern GPUs.

> There were 42 million transistors in a Pentium 4

USB 1.1 was in '96

That's Pentium 1 era. Pentium 2 was released year after. Pentium 1 had one tenth of that

> When USB came out we already had some very crude 3D realtime rendering already at that time available to mass consumers, I'm sure we could have figured out how to flip power on a port if we put our minds to it.

Why your six neuron brain can't imagine saving pennies on chips used in cheap computer peripherals?

Reply on news.ycombinator.com